Self-Improvement in LLMs

2 min read

Self-improvement methods allow LLMs to improve their own capabilities by generating and learning from their own outputs.

STaR (Self-Taught Reasoner, Zelikman et al., 2022):

  1. Model attempts to solve problems with chain-of-thought
  2. Keep solutions that arrive at correct answers
  3. Fine-tune the model on its own correct reasoning traces
  4. Repeat — the model bootstraps better reasoning over iterations

Iterative DPO:

  1. Generate response pairs from the current model
  2. Score/rank them (with a reward model or verifier)
  3. Train with DPO on the self-generated preferences
  4. Repeat with the improved model

When self-improvement works:

  • There's a reliable verifier (math, code — can check correctness)
  • The model can occasionally produce correct solutions (it has the latent capability)
  • Each iteration filters for quality and amplifies good reasoning patterns

When it collapses:

  • Without a reliable verifier, the model reinforces its own errors (mode collapse)
  • The model may converge to narrow strategies that exploit the verifier
  • Diversity decreases across iterations if not managed
  • Reward hacking: if the verifier is imperfect, the model learns to fool it

DeepSeek-R1: large-scale demonstration of self-improvement — RL on self-generated reasoning traces, bootstrapped from a small seed of human demonstrations.

See also: Chain-of-Thought Prompting, Test-Time Compute, Direct Preference Optimization

Linked from